Quick & dirty deployments on Ec2 with Python, Django, Pip, and Git.

As I was working on our wedding website, I quickly started realizing my fiancĂ© and others would need to view it as it went up. And make changes… lots of them.

The idea of one click (key) deploys is becoming something of a religion for me with my hobby projects so why was it any different for our wedding site… so I decided to start from complete scratch and put together an Ubuntu 10.10 server on EC2 to the task of hosting my site with a one key deploy using Git, document the steps, and share them in case it helps you in your efforts. Sure I could’ve used localtunnel or something like it, but where’s the fun in that. I want to try this app live in the same server I plan to deploy it’s beta in, and learn something along the way.

Disclaimer: There are lots of great ways to automate deployments, especially with Python and OSS. Among or perhaps the most notable is Fabric, which is built for just this challenge. I don’t know Fabric well enough to make a simple tutorial of it —- but I do know Git, and I wanted something really simple.

Assumptions: Your app is ready to be deployed. It works locally on the django development server. You know the basics of POSIX, AWS & EC2, Python, Django, Git, and are running a reasonably current Mac or POSIX OS, and have done at least some of this before.

As a simplifying convention, I’ll use “LOCAL MACHINE” followed by steps to be done on your local dev machine, and “REMOTE MACHINE” followed by steps to be done on your remote EC2 instance. In case it’s not clear don’t type anything in <note:> brackets.

Let’s begin…

LOCAL MACHINE

  • Launch a new server instance, in this example: Ubuntu 10.10 via AMI: ami-ccf405a5
  •  Enable SSH, HTTP, HTTPS (and anything else you want)
  •  Create & save (or re-use) your key locally so we can further refer to it as <awsKey>, typically on local path ~/.ssh/<awsKey>

Note: Elastic IPs are nice to use here. They just make life easier for a number of reasons, and particularly if you’re going to put this on a domain later it can be handy depending on which DNS route you choose. In our case we’ll use so our local hosts file (POSIX) can point reliably to a named server to shorten shell incantations…LOCAL MACHINE

  • Edit your hosts file (on path /etc/hosts) to append a line (hopefully replacing the <elastic-ip….> part with the actual ip):

    <elastic-ip.to.remote.server>ec2server

  • Login to the remote machine, “ec2server”

    $ ssh -i ~/.ssh/<aws_key> ubuntu@ec2server

Now it’s time to get some of the basics out of the way and install git…

REMOTE MACHINE

$ sudo apt-get update && sudo apt-get upgrade $ sudo dpkg-reconfigure tzdata <note: to set timezone> $ sudo apt-get install git-core git

Now that the server has the absolute minimum for this exercise we can create our git repository and keep pushing forward. Note here that if you plan to use this for more public facing deployments it’s best to secure the machine at this stage.

$ cd ~/ && mkdir productionsite.git && cd productionsite.git $ git init —bare

Now that the repo is initialized, we’ll be able to add it as a remote on our local dev machine. The naturally obvious thing to note here is that just publishing code to the remote Git repo does nothing to deploy it. For the deployment steps, we’ll use a post-receive hook on the repo. In short in case it’s new information for you, Git automatically ships with a handful of hooks which are simply shell scripts that Git executes when certain actions occur. By creating / editing these scripts (hooks) you can do all kinds of neat things as you take action with the system. For our purposes, we want to simply checkout all the code from the repo into a directory on the remote machine which we can then run a django development server on. All of this can be automated, but for now, let’s just checkout the code and we’ll manually run the development server. 

Continuing on, make a directory to hold the checked-out code from the repo

$ cd /home/ubuntu $ mkdir web

Now that we have our destination for checked out code to /home/ubuntu/web let’s make a quick post-receive hook that does the actions described above. It’s 2 lines…

$ cd /home/ubuntu/productionsite.git $ cat > hooks/post-receive #!/bin/sh GIT_WORK_TREE=/home/ubuntu/web git checkout -f <note: ctrl+d finishes this command>

$ chmod +x hooks/post-receive

That is all there is too it - pretty straightforward right? Now it’s time to go back into your local dev machine and connect your local repo to the new remote, and test everything out. At this point, I’m assuming you have a local project already setup somewhere, possibly already remoted to GitHub or something - at least with Git already initialized.

LOCAL MACHINE

$ cd <path-to-local-project-directory> $ ssh-add ~/.ssh/<aws_key> $ git remote add ec2 ssh://ubuntu@ec2server/home/ubuntu/productionsite.git

Now your remote is named “ec2” and points to the remote machine named “ec2server”. 

On your first push, you will need to setup the repo with the right head information so the first push should be:

$ git push ec2 +master:refs/heads/master

Subsequent pushes you can simply use:

$ git push ec2

Note here that after your first push, your code is published to your remote machine and should be checked out into the /home/ubuntu/web directory that we setup for it. So let’s login and do some validation that the push and post-receive hook work as expected.

 REMOTE MACHINE

$ cd /home/ubuntu/web && ls <note: you should see your code checked out here>

If your code isn’t checked out, a good first place to double check is back in the shell on your local machine. Git will print error messages it received from your ec2 remote repo on the local shell window as a part of the push notes. 

Now you’re done with nearly the simplest possible code deployment setup. You can deploy in one command, but now you need django and all the key software components of your app installed on the server. The reason I install python dev tools at this step instead of at the beginning is to re-enforce the fact that this is a loosely coupled exercise. You really can stay quick & dirty with Git remotes, and use any other tools from there out.

Here forward, I’m assuming you used distribute and did a “$ pip freeze > requirements.txt” to capture all the packges you need. I’ll also assume that requirements.txt is in your django project folder and got pushed to your remote server successfully as a result of the previous steps in this exercise. If you didn’t you want to run the following on your local machine in your django project directory:

$ pip freeze > requirements.txt $ git add requirements.txt && git commit -am ‘adding requirements.txt’ && git push ec2

Let’s move on and install the dependencies on our remote system.

REMOTE MACHINE

$ sudo apt-get install python-pip python-dev build-essential $ sudo pip install —upgrade pip$ pip freeze <note: to see what’s already there and validate the toolchain’s working> $ sudo pip install -r /home/ubuntu/web/<rest-of-path-to-django-proj-dir>/requirements.txt

That’s it, you should have your apps environment installed. We didn’t use a virtualenv on the server because we’re lazy in this tutorial and probably it could be argued reasonably easily that there’s no need when you have a dedicated server for your app. Either way, now it’s time to make sure your app works.

At this point you might have some settings in your settings.py file that are wired to directories on your development machine, or other nuances. It’s a good time to open your settings.py file and make sure you’re ready for the remote machine. Sometimes the easiest way to do this is start the development webserver and take a look at your app - the same command you can use when you’re all done and ready to show off your work as it’s being built!

$ cd /home/ubuntu/web <note: assuming this is where your manage.py is> $ sudo python manage.py runserver 0.0.0.0:80 <note: to run the django development server on port 80>

All done! Hope that’s helpful, feedback welcome.